by Wan, Kwok Fai. / Thesis (Ph.D.)--Chinese University of Hong Kong, 1995. / Includes bibliographical references (leaves 138-148). / Chapter Chapter 1. --- Adaptive signal processing and its applications --- p.1 / Chapter 1.1. --- Introduction --- p.1 / Chapter 1.2. --- Applications of adaptive system --- p.3 / Chapter 1.2.1. --- Adaptive noise cancellation --- p.3 / Chapter 1.2.2. --- Adaptive echo cancellation --- p.5 / Chapter 1.2.3. --- Adaptive line enhancement --- p.5 / Chapter 1.2.4. --- Adaptive linear prediction --- p.7 / Chapter 1.2.5. --- Adaptive system identification --- p.8 / Chapter 1.3. --- Algorithms for adaptive systems --- p.10 / Chapter 1.4. --- Transform domain adaptive filtering --- p.12 / Chapter 1.5 --- The motivation and organization of the thesis --- p.13 / Chapter Chapter 2. --- Time domain split-path adaptive filter --- p.16 / Chapter 2.1. --- Adaptive transversal filter and the LMS algorithm --- p.17 / Chapter 2.1.1. --- Wiener-Hopf solution --- p.17 / Chapter 2.1.2. --- The LMS adaptive algorithm --- p.20 / Chapter 2.2. --- Split structure adaptive filtering --- p.23 / Chapter 2.2.1. --- Split structure of an adaptive filter --- p.24 / Chapter 2.2.2. --- Split-path structure for a non-symmetric adaptive filter --- p.25 / Chapter 2.3. --- Split-path adaptive median filtering --- p.29 / Chapter 2.3.1. --- Median filtering and median LMS algorithm --- p.29 / Chapter 2.3.2. --- The split-path median LMS (SPMLMS) algorithm --- p.32 / Chapter 2.3.3. --- Convergence analysis of SPMLMS --- p.36 / Chapter 2.4. --- Computer simulation examples --- p.41 / Chapter 2.5. --- Summary --- p.45 / Chapter Chapter 3. --- Multi-stage split structure adaptive filtering --- p.46 / Chapter 3.1. --- Introduction --- p.46 / Chapter 3.2. --- Split structure for a symmetric or an anti-symmetric adaptive filter --- p.48 / Chapter 3.3. --- Multi-stage split structure for an FIR adaptive filter --- p.56 / Chapter 3.4. --- Properties of the split structure LMS algorithm --- p.59 / Chapter 3.5. --- Full split-path adaptive algorithm for system identification --- p.66 / Chapter 3.6. --- Summary --- p.71 / Chapter Chapter 4. --- Transform domain split-path adaptive algorithms --- p.72 / Chapter 4.1. --- Introduction --- p.73 / Chapter 4.2. --- general description of transforms --- p.74 / Chapter 4.2.1. --- Fast Karhunen-Loeve transform --- p.75 / Chapter 4.2.2. --- Symmetric cosine transform --- p.77 / Chapter 4.2.3. --- Discrete sine transform --- p.77 / Chapter 4.2.4. --- Discrete cosine transform --- p.78 / Chapter 4.2.5. --- Discrete Hartley transform --- p.78 / Chapter 4.2.6. --- Discrete Walsh transform --- p.79 / Chapter 4.3. --- Transform domain adaptive filters --- p.80 / Chapter 4.3.1. --- Structure of transform domain adaptive filters --- p.80 / Chapter 4.3.2. --- Properties of transform domain adaptive filters --- p.83 / Chapter 4.4. --- Transform domain split-path LMS adaptive predictor --- p.84 / Chapter 4.5. --- Performance analysis of the TRSPAF --- p.93 / Chapter 4.5.1. --- Optimum Wiener solution --- p.93 / Chapter 4.5.2. --- Steady state MSE and convergence speed --- p.94 / Chapter 4.6. --- Computer simulation examples --- p.96 / Chapter 4.7. --- Summary --- p.100 / Chapter Chapter 5. --- Tracking optimal convergence factor for transform domain split-path adaptive algorithm --- p.101 / Chapter 5.1. --- Introduction --- p.102 / Chapter 5.2. --- The optimal convergence factors of TRSPAF --- p.104 / Chapter 5.3. --- Tracking optimal convergence factors for TRSPAF --- p.110 / Chapter 5.3.1. --- Tracking optimal convergence factor for gradient-based algorithms --- p.111 / Chapter 5.3.2. --- Tracking optimal convergence factors for LMS algorithm --- p.112 / Chapter 5.4. --- Comparison of optimal convergence factor tracking method with self-orthogonalizing method --- p.114 / Chapter 5.5. --- Computer simulation results --- p.116 / Chapter 5.6. --- Summary --- p.121 / Chapter Chapter 6. --- A unification between split-path adaptive filtering and discrete Walsh transform adaptation --- p.122 / Chapter 6.1. --- Introduction --- p.122 / Chapter 6.2. --- A new ordering of the Walsh functions --- p.124 / Chapter 6.3. --- Relationship between SM-ordered Walsh function and other Walsh functions --- p.126 / Chapter 6.4. --- Computer simulation results --- p.132 / Chapter 6.5. --- Summary --- p.134 / Chapter Chapter 7. --- Conclusion --- p.135 / References --- p.138

No campo do processamento digital de sinais e em especial da filtragem adaptativa, procura-se continuamente algoritmos que sejam rápidos e simples. Neste contexto, este trabalho apresenta o estudo de novos algoritmos de tempo discreto denominados algoritmos aceleradores (completo, regressivo e progressivo), obtidos a partir da discretização de um algoritmo de tempo contínuo baseado no ajuste da segunda derivada (aceleração) da estimativa dos parâmetros. Destes algoritmos optou-se por estudar mais aprofundadamente os algoritmos aceleradores progressivo e regressivo, devido respectivamente a sua menor complexidade computacional e ao seu desempenho. Para este estudo e análise foram escolhidos como base de comparação os algoritmos LMS e NLMS. Isto porque estes algoritmos estão entre os mais usados e, assim como os algoritmos aceleradores, podem ser obtidos a partir da discretização de algoritmos de tempo contínuo através dos métodos de Euler progressivo e regressivo respectivamente. A análise do algoritmo progressivo mostrou que seu desempenho é inferior ao do algoritmo LMS. Visando diminuir a complexidade computacional do algoritmo acelerador regressivo, foi obtido um novo algoritmo: o versão g. Assim a análise focou-se no algoritmo acelerador regressivo versão g, o qual apresentou um desempenho bom quando comparado no desajuste e no tracking com o algoritmo NLMS, mostrando um melhor compromisso entre velocidade de convergência e variância das estimativas. Este bom desempenho foi comprovado por análises teóricas, por simulações e através da aplicação deste algoritmo na equalização de um canal variante no tempo. / In the digital signal processing field and specially in adaptive filtering, there is a constant search for algorithms both simple and with good performance. This work presents new discrete-time algorithms called accelerating algorithms (APCM and ARg), obtained through the discretization of a continuous-time algorithm that uses the second derivate (acceleration) to adjust the parameter estimates. We provide theoretical analyses for both algorithms, finding expressions for the mean and mean-square errors in the parameter estimates. In addition, we compare the performance of the accelerating algorithms with LMS and NLMS. The analysis of the APCM algorithm showed that its performance is inferior to that of the LMS algorithm. On the other hand, the ARg algorithm presented good performance when compared in terms of misadjustment and tracking with the NLMS algorithm, showing a better compromise between convergence speed and variance of the estimates. This better performance was proven by theoretical analyses, by simulations and through the application of this algorithm to the equalization of a time-variant channel.

The advantages of administering an adaptive test battery, a collection of multiple adaptive subtests that are specifically tailored to examinees’ abilities, include shortening the subtest length and maintaining the accuracy of individual subtest scores. The test battery can incorporate a range of subjects, though this study focused primarily on Math and Reading. This study compared different ways of incorporating collateral information (CI), supplementary information beyond examinees’ current test performance, under two frameworks (Unidimensional and Multidimensional computerized adaptive testing). It also investigated the impact of subtest intercorrelations (the relationship between an examinee’s test scores), as well as the sequences of subtest administration on ability estimation in a variable-length adaptive battery. Practical issues including content constraints and item exposure control were also considered. Findings showed that the CI methods improved measurement efficiency with an acceptable level of measurement precision. The CI was more beneficial when associated with higher intercorrelations among the subtests. Also, the CI was found to be advantageous during the early stages of the subtests which were not taken first. Therefore, the CI may improve the examinee experience by administering items more aligned with their abilities. In addition, the CI should reduce costs for testing organizations by requiring fewer items and possibly saving seat time, while still providing reliable scores. The results should help practitioners decide whether the use of the CI is worthwhile under their particular testing situation.

The thesis deals with the design and implementation of an Adaptive Flight Control technique for Unmanned Aerial Vehicles (UAVs). The application of UAVs has been increasing exponentially in the last decade both in Military and Civilian fronts. These UAVs fly at very low speeds and Reynolds numbers, have nonlinear coupling, and tend to exhibit time varying characteristics. In addition, due to the variety of missions, they fly in uncertain environments exposing themselves to unpredictable external disturbances. The successful completion of the UAV missions is largely dependent on the accuracy of the control provided by the flight controllers. Thus there is a necessity for accurate and robust flight controllers. These controllers should be able to adapt to the changes in the dynamics due to internal and external changes. From the available literature, it is known that, one of the better suited adaptive controllers is the model based controller. The design and implementation of model based adaptive controller is discussed in the thesis. A critical issue in the design and application of model based control is the online identification of the UAV dynamics from the available sensors using the onboard processing capability. For this, proper instrumentation in terms of sensors and avionics for two platforms developed at UNSW@ADFA is discussed. Using the flight data from the remotely flown platforms, state space identification and fuzzy identification are developed to mimic the UAV dynamics. Real time validations using Hardware in Loop (HIL) simulations show that both the methods are feasible for control. A finer comparison showed that the accuracy of identification using fuzzy systems is better than the state space technique. The flight tests with real time online identification confirmed the feasibility of fuzzy identification for intelligent control. Hence two adaptive controllers based on the fuzzy identification are developed. The first adaptive controller is a hybrid indirect adaptive controller that utilises the model sensitivity in addition to output error for adaptation. The feedback of the model sensitivity function to adapt the parameters of the controller is shown to have beneficial effects, both in terms of convergence and accuracy. HIL simulations applied to the control of roll stabilised pitch autopilot for a typical UAV demonstrate the improvements compared to the direct adaptive controller. Next a novel fuzzy model based inversion controller is presented. The analytical approximate inversion proposed in this thesis does not increase the computational effort. The comparisons of this controller with other controller for a benchmark problem are presented using numerical simulations. The results bring out the superiority of this technique over other techniques. The extension of the analytical inversion based controller for multiple input multiple output problem is presented for the design of roll stabilised pitch autopilot for a UAV. The results of the HIL simulations are discussed for a typical UAV. Finally, flight test results for angle of attack control of one of the UAV platforms at UNSW@ADFA are presented. The flight test results show that the adaptive controller is capable of controlling the UAV suitably in a real environment, demonstrating its robustness characteristics.

This thesis investigates techniques and algorithms for mitigating radio frequency interference (RFI) affecting radio astronomy observations. In the past radio astronomy has generally been performed in radio-quiet geographical locations and unused parts of the radio spectrum, including small protected frequency bands. The increasing use of the entire spectrum and global transmitters such as satellites are forcing the astronomy community to begin implementing active interference cancelling. The amount of harmful interference affecting observations will also increase as future instruments such as the Square Kilometre Array (SKA) are required to use larger bandwidths to reach up to 100 times the current sensitivity levels, and as spectral line observations require observing in bands licensed to other spectrum users. Particular attention is paid to interference cancellation algorithms which make use of reference beams. This has proven to be successful in removing interference from the contaminated astronomical data. Reference antenna cancellers are closely analysed, leading to filters and techniques that can offer improved RFI excision for some important applications. It is shown that pre- and post-correlation reference antenna cancellers give similar results, and an important aspect of the cancellers is the use of a second reference signal when the reference interference-to-noise ratio is low. These modified filters can theoretically offer infinite interference suppression in the voltage domain, equivalent to that of post-correlation interference cancellers, and their internal structure can offer an understanding of the residual RFI and added receiver noise components of a variety of reference antenna techniques. The effect of variable geometric delays is also considered and various filters are compared as a function of the geometric fringe rate.

One objective of artificial intelligence is to model the behavior of an intelligent agent interacting with its environment. The environment's transformations can be modeled as a Markov chain, whose state is partially observable to the agent and affected by its actions; such processes are known as partially observable Markov decision processes (POMDPs). While the environment's dynamics are assumed to obey certain rules, the agent does not know them and must learn. In this dissertation we focus on the agent's adaptation as captured by the reinforcement learning framework. This means learning a policy---a mapping of observations into actions---based on feedback from the environment. The learning can be viewed as browsing a set of policies while evaluating them by trial through interaction with the environment. The set of policies is constrained by the architecture of the agent's controller. POMDPs require a controller to have a memory. We investigate controllers with memory, including controllers with external memory, finite state controllers and distributed controllers for multi-agent systems. For these various controllers we work out the details of the algorithms which learn by ascending the gradient of expected cumulative reinforcement. Building on statistical learning theory and experiment design theory, a policy evaluation algorithm is developed for the case of experience re-use. We address the question of sufficient experience for uniform convergence of policy evaluation and obtain sample complexity bounds for various estimators. Finally, we demonstrate the performance of the proposed algorithms on several domains, the most complex of which is simulated adaptive packet routing in a telecommunication network.

The complexity of software systems is exploding, along with their use and application in new domains. Managing this complexity has become a focal point for research in Software Engineering. One direction for research in this area is developing techniques for designing adaptive software systems that self-optimize, self-repair, self-configure and self-protect, thereby reducing maintenance costs, while improving quality of service. This thesis presents a requirements-driven approach for developing adaptive and customizable systems. Requirements goal models are used as a basis for capturing problem variability, leading to software designs that support a space of possible behaviours – all delivering the same functionality. This space can be exploited at system deployment time to customize the system on the basis of user preferences. It can also be used at runtime to support system adaptation if the current behaviour of the running system is deemed to be unsatisfactory. The contributions of the thesis include a framework for systematically generating designs from high-variability goal models. Three complementary design views are generated: configurational view (feature model), behavioural view (statecharts) and an architectural view (parameterized architecture). The framework is also applied to the field of business process management for intuitive high-level process customization. In addition, the thesis proposes a modeling framework for capturing domain variability through contexts and applies it to goal models. A single goal model is used to capture requirements variations in different contexts. Models for particular contexts can then be automatically generated from this global requirements model. As well, the thesis proposes a new class of requirements-about-requirements called awareness requirements. Awareness requirements are naturally operationalized through feedback controllers – the core mechanisms of every adaptive system. The thesis presents an approach for systematically designing monitoring, analysis/diagnosis, and compensation components of a feedback controller, given a set of awareness requirements. Situations requiring adaptation are explicitly captured using contexts.

The complexity of software systems is exploding, along with their use and application in new domains. Managing this complexity has become a focal point for research in Software Engineering. One direction for research in this area is developing techniques for designing adaptive software systems that self-optimize, self-repair, self-configure and self-protect, thereby reducing maintenance costs, while improving quality of service. This thesis presents a requirements-driven approach for developing adaptive and customizable systems. Requirements goal models are used as a basis for capturing problem variability, leading to software designs that support a space of possible behaviours – all delivering the same functionality. This space can be exploited at system deployment time to customize the system on the basis of user preferences. It can also be used at runtime to support system adaptation if the current behaviour of the running system is deemed to be unsatisfactory. The contributions of the thesis include a framework for systematically generating designs from high-variability goal models. Three complementary design views are generated: configurational view (feature model), behavioural view (statecharts) and an architectural view (parameterized architecture). The framework is also applied to the field of business process management for intuitive high-level process customization. In addition, the thesis proposes a modeling framework for capturing domain variability through contexts and applies it to goal models. A single goal model is used to capture requirements variations in different contexts. Models for particular contexts can then be automatically generated from this global requirements model. As well, the thesis proposes a new class of requirements-about-requirements called awareness requirements. Awareness requirements are naturally operationalized through feedback controllers – the core mechanisms of every adaptive system. The thesis presents an approach for systematically designing monitoring, analysis/diagnosis, and compensation components of a feedback controller, given a set of awareness requirements. Situations requiring adaptation are explicitly captured using contexts.

Thesis (Ph.D.)--University of Central Florida, 2008. / Adviser: Wasfy B. Mikhael. Includes bibliographical references (p. 152-166).

The existing powertrain control system in Volvo CE's vehicles consists of various types of physical quantities that are controlled. One of them is the engine speed. The purpose of this thesis is to investigate whether there are other control strategies suitable for engine speed control, than the existing one. Currently, the existing control system requires re-calibration of the control parameters if hardware in the vehicle is replaced. The current controller is a gain-scheduled PID controller with control parameters that varies over the operating range. The aim has been to develop several different adaptive control strategies. Adaptive control methods are expected to adapt to the changes of the system that a replacement of hardware can bring. The performance and robustness of the developed controllers have been compared with the existing controller. The approach has been to implement the control strategies in Matlab/Simulink and simulate the process with existing engine software provided by Volvo CE. The next step was to test and verify the controllers in a real machine. The focus in this thesis work has been on the adaptive control strategies MRAC (Model-Reference Adaptive Control) and L1 Adaptive Control. In the MRAC structure the desired performance is specified in terms of a reference model that the real system is supposed to follow. Each time an error is generated, by comparing actual and desired output, a suitable algorithm is used in order to obtain the control signal that can minimize the error. In addition, modeling errors and disturbances are estimated so that the controller can compensate for these. L1 Adaptive Control is an extension of the MRAC structure. The difference is that before the control signal is fed to the real system, it is low-pass filtered. This is done in order to prevent feeding high frequencies into the system. The results show that adaptive control has potential to be used in engine speed control. Reference following and disturbance rejection is well handled and simulations have furthermore shown that the developed controllers can deal with changes in the hardware. One of the developed L1-controllers was implemented in a real machine with promising results. / Det existerande styrsystemet i Volvo CE:s maskiner har till uppgift att styra och reglera flera olika fysikaliska storheter. En av dessa storheter är motorvarvtalet. Syftet med detta examensarbete är att undersöka alternativa reglerstrategier som kan användas för att styra motorvarvtalet. Problemet idag är att det nuvarande styrsystemet kräver omkalibrering av regulatorparametrar när befintlig hårdvara i maskinen behöver ersättas på grund av föråldring eller slitage. Den nuvarande regulatorn är en parameterstyrd PID-regulator där regulatorparametrarna beror av aktuell arbetspunkt. Målet har varit att utveckla och prova flera olika adaptiva reglerstrategier. Dessa metoder förväntas kunna hantera förändringar och adaptera sig mot nya förhållanden och omständigheter som en hårdvaruförändring kan medföra. Prestanda och robusthet som de utvecklade regulatorerna erhåller har jämförts mot den existerande regulatorstrukturen. Tillvägagångssättet har varit att implementera reglerstrategierna i Matlab/Simulink samt simulera med tillhörande motormjukvara som Volvo CE tillhandahållit. I nästa fas skulle regulatorerna testas och verifieras i en riktig maskin. Fokuset har under detta examensarbete riktats mot de två adaptiva reglerstrategierna Model-Reference Adaptive Control (MRAC) och L1 Adaptive Control. MRAC-strukturen bygger på att specificera prestandan genom en referens-modell som det riktiga systemet skall följa. Varje gång en avvikelse uppstår så beräknas en lämplig styrsignal genom att beakta och försöka minimera skillnaden mellan det riktiga systemet och den önskade referens-modellen. Till detta modelleras och skattas störningar som regulatorn skall kompensera för. Tekniken inom L1 Adaptive Control är en utvidgning av MRAC. Önskat beteende specificeras även för denna regulatorstruktur men största skillnaden är att innan styrsignalen matas in till systemet så lågpassfiltreras den. Detta görs i förebyggande syfte för att inte släppa in onödigt höga frekvenser in i systemet. Resultaten visar att adaptiv reglering av motorvarvtalet har potential. Referensföljning och undertryckning av störningar hanteras väl och simuleringar har dessutom visat att de utvecklade regulatorerna kan hantera hårdvaruändringar. En av de utvecklade L1-regulatorerna implementerades i en riktig maskin och resultaten såg lovande ut.